The Situation

As the organization grew, so did the volume of routine HR questions — attendance records, leave balances, parental leave entitlements, payroll details, and policy clarifications. These were handled through email, ticketing systems, or direct conversations, and the source material describes HR business partners spending "a disproportionate share of their time" on low-complexity queries rather than strategic workforce initiatives.

The same fragmentation showed up for people managers, who needed to track attendance and leave patterns across rolling six- and twelve-month periods — data that existed in the HR system but wasn't easily surfaced in a query-able form. New joiners faced the sharpest version of the problem: onboarding required up to ten hours of guided training just to learn where to find internal HR information.

The Approach

The team built the HR agent as a conversational tool embedded directly in the client's internal chat platform, backed by a Retrieval-Augmented Generation architecture and a live-data integration layer for querying HR systems in real time.

Two design choices carry the weight of the case: personal-scope, credential-based access — each employee authenticates with enterprise credentials, and the agent only returns data that person is authorized to see, enforced in application code rather than left to prompt instructions — and mandatory source citation, where every answer ends with a reference to the specific policy document, section, or data range it drew from.

Rollout was phased rather than all at once. Phase 1, live at time of writing, is read-only: answering queries, surfacing attendance and leave data for managers, and providing policy guidance. Phase 2, planned but not yet built, extends the agent to raising leave requests, escalating issues, and tracking them to resolution — moving from an information layer to a transactional one.

The Results

What the material does support qualitatively: the citation model was treated as a compliance requirement rather than a feature. Because policy accuracy touches payroll and employee rights directly, the ability to show a verifiable source on every answer is described as "a baseline requirement for trust," not a differentiator. Privacy guardrails were likewise hardcoded into the application logic — not delegated to the prompt — specifically so that no amount of prompt manipulation could expose one employee's data to another.

Key Takeaways

• Enforce data boundaries in code, not in the prompt. A credential-scoped access model survives adversarial input in a way that prompt-level instructions don't.
• In a domain where accuracy has payroll or compliance consequences, treat source citation as infrastructure, not a nice-to-have.
• Ship the read-only version first. Establishing trust in the agent's answers before it can take action on someone's behalf lowers the risk of the transactional phase.
• Design the first use case as one module of a shared platform, not a one-off build — it's what made extending the same agent into other departments faster later.